Analog to Digital Conversion: What It Is, Types, and Specs

Analog to digital conversion (often written as analogue‑to‑digital conversion, or ADC) is how we turn real‑world signals into numbers a computer can store, edit and share. A microphone’s changing voltage becomes an audio file, the picture on a VHS tape becomes pixels, a temperature sensor’s output becomes a reading. At its heart, an ADC takes snapshots of a continuous signal, rounds each snapshot to the nearest available level, and encodes those levels as binary values.

📞 Don’t delay, send your enquiry right now using the form below or phone 0437 822 357.

This guide explains how ADCs work in practice—sampling, quantisation and encoding—then unpacks the specs that matter, like sampling rate, Nyquist, aliasing, resolution and step size. You’ll learn what SNR, ENOB, linearity and distortion really mean, when to choose flash, SAR, delta‑sigma or pipeline architectures, and how timing, jitter and latency affect results. We’ll cover analogue front‑ends, references, digital interfaces and output coding, with practical microcontroller examples (e.g., Arduino) and clear advice for audio, video and film digitisation. Let’s begin with the core conversion process.

How analogue-to-digital conversion works

Think of an ADC like pausing a moving scene: it briefly holds the analogue level, rounds it to the nearest permitted value, then writes that value as a number. This happens over and over, fast enough that the stream of numbers closely tracks the original signal. In practice, an ADC uses a sampling stage (often with a sample‑and‑hold) to capture the instantaneous voltage, a quantisation stage to choose the nearest level from a set of 2^n possibilities within the reference range, and an encoding stage to output the result as binary.

Sample: The input is momentarily “frozen” so it can be measured cleanly.
Quantise: The held value is mapped to the nearest code level, introducing a small rounding (quantisation) error.
Encode: The chosen level is formatted as digital data and sent to the host.

These steps—sampling, quantisation and encoding—are the common core across ADC types. Next, we’ll decide how fast to sample and why Nyquist matters.

Sampling rate, Nyquist and aliasing

Sampling rate defines how often you measure the signal: fs samples per second, with fs = 1/T where T is the sampling period. To avoid losing information, the Nyquist criterion requires the sampling frequency to be at least twice the highest frequency present: fs >= 2 * fmax. Undersampling breaks this rule and causes aliasing—high‑frequency content folds back as false low‑frequency components—so your digital data no longer represents the original analogue signal. Because real signals carry noise and harmonics, analogue to digital conversion almost always needs an anti‑aliasing low‑pass filter ahead of the ADC to limit bandwidth to below fs/2 and prevent unexpected high‑frequency content from slipping through.

Basic rule: Identify fmax and choose fs ≥ 2 * fmax (e.g., if fmax = 100 kHz, pick fs ≥ 200 kS/s).
Use filtering: Add a pre‑ADC low‑pass filter with a cutoff comfortably below fs/2 to control aliasing.
Be pragmatic: For slow sensors, modest fs is fine; for fast or complex spectra, add margin above Nyquist for safer filtering.

Resolution, bit depth and step size

Resolution (bit depth) tells you how many distinct levels an ADC can report. With n bits there are 2^n codes, so the analogue-to-digital step size—often called one least significant bit (LSB)—is LSB = Vref / 2^n. This is the smallest voltage change the converter can meaningfully represent. For example, a 12‑bit ADC with Vref = 5 V has 4096 levels, so LSB ≈ 5 / 4096 ≈ 1.22 mV—fine enough to track small changes. Conversely, a 10‑bit Arduino‑class ADC yields 1024 levels across its reference, mapping readings from 0 to 1023.

Bit depth doesn’t make your signal cleaner; it determines the grid you round to during quantisation. Two practical levers matter: the number of bits and the reference range you choose. Match Vref to the expected input swing to use as many codes as possible without clipping.

More bits, smaller steps: Every extra bit halves LSB.
Set Vref sensibly: A tighter Vref around your signal maximises usable resolution.
Ratiometric behaviour: With Vref = VDD, results scale with supply, which can be useful for ratio sensors.

Performance metrics: SNR, ENOB, linearity and distortion

Beyond speed and resolution, analogue‑to‑digital conversion quality depends on how noisy, linear and distortion‑free the conversion is. Data sheets summarise this with a few core metrics. Read them correctly and you can predict whether an ADC preserves quiet details, keeps levels accurate across the range, and avoids spurious tones that contaminate spectra.

SNR (signal‑to‑noise ratio): Ratio of the wanted tone to noise within the measurement bandwidth, excluding distortion. Higher SNR means cleaner low‑level detail and less hiss.
ENOB (effective number of bits): Real‑world usable resolution derived from measured noise/distortion (often via SINAD). Compare ENOB at the same bandwidth and output rate.
Linearity (INL/DNL): DNL reflects step‑to‑step uniformity; excessive DNL risks missing codes. INL captures transfer‑curve bowing that shifts levels and seeds harmonic error; good monotonicity is essential.
Distortion (THD/SFDR): THD sums harmonic energy; SFDR is the gap between the fundamental and the largest spur. Low THD/high SFDR protect dynamic range in audio and RF work.

Core ADC architectures and when to use each

Different ADC architectures trade speed, resolution, latency and power in distinct ways. Choosing the right core is about matching your signal bandwidth and accuracy needs to the converter’s strengths, and understanding any front‑end filtering or digital processing the design will require.

Flash (parallel): Compares the input against many thresholds at once using 2^n – 1 comparators. Fastest conversion with minimal latency, but typically lower resolution and higher power. Use for very high‑speed capture, triggers and RF/oscilloscope front ends.
SAR (successive approximation): Uses a binary search with a DAC and comparator after a sample‑and‑hold. Excellent general‑purpose choice with strong resolution–speed balance and low power. Ideal for sensors, control loops, data acquisition and microcontrollers.
Delta‑sigma (sigma‑delta): Oversamples and noise‑shapes the input, then digitally filters/decimates. Delivers very high resolution and linearity with built‑in filtering, at the cost of latency and lower signal bandwidth. Best for precision audio, temperature/pressure, weigh scales and instrumentation.
Pipeline: Processes the conversion in stages (coarse‑to‑fine). Achieves very high sample rates with moderate resolution and fixed latency. Suits IF sampling, communications, radar and high‑speed instrumentation or video.
Integrating (dual‑slope): Integrates over time to average noise; extremely accurate for DC/low‑frequency, but slow. Common in digital multimeters and energy metering.
Hybrids/subranging: Combine techniques (e.g., SAR front end with digital filtering) to blend speed and precision when requirements overlap.

Timing, latency and jitter

Timing determines when a converter samples and when its result appears. Throughput is limited by the sample‑and‑hold acquisition time plus the conversion time; if either is too slow, codes smear or arrive late. Latency is the delay between an analogue event and the corresponding digital code. SAR converters have near‑instantaneous, cycle‑by‑cycle results; pipeline and delta‑sigma devices introduce fixed latency (often several sample periods) because of staged conversion and digital filtering. Jitter is short‑term variation of the sampling instant (from the clock or aperture) that turns into noise; its impact grows with input frequency, so fast audio, IF and video signals are especially sensitive.

Know your budget: Add conversion, filtering and interface delays to meet control‑loop or A/V sync requirements.
Match clocks to bandwidth: Use a low‑phase‑noise, low‑jitter clock for high‑frequency inputs.
Allow acquisition time: Drive sources with low impedance or a buffer so the S/H can settle each cycle.
Fix the pipeline: In multi‑stage or decimating ADCs, account for fixed group delay in processing chains.

Analogue front-end, inputs and references

Before an ADC can shine, the analogue front‑end (AFE) must prepare the signal. It filters out‑of‑band energy, scales and biases voltages into the allowed input range, presents a low impedance to the sample‑and‑hold, and protects the converter. Just as crucial, the reference pins set full scale (FS) and the LSB grid; noise or drift on Vref turns straight into conversion error. Treat the AFE and reference as one accuracy budget.

Buffer and filter: Low‑noise buffer plus anti‑alias low‑pass; set cutoff comfortably below fs/2.
Source impedance: SAR sampling caps need drive; keep resistance low or use an op‑amp driver.
Input topology: Single‑ended is simple; differential improves CMRR/linearity; set VCM in range.
Scale and protect: Dividers/AC‑coupling, small series R and clamps/TVS to meet limits.
Reference: Use low‑noise, low‑drift Vref or ratiometric; decouple at REF pins; route quietly.
Layout/grounds: Short input/reference traces; separate analogue/digital returns; keep clocks away.

Digital interfaces and output coding

After analogue to digital conversion, the result must be clocked out predictably. Most ADCs present either a synchronous serial stream (low pin‑count, ideal for microcontrollers) or a wider, time‑aligned bus for higher throughputs. A data‑ready or output clock pin defines when each bit or word is valid, and devices often let you select bit order, word length and how multi‑channel frames are packed so your host can capture cleanly.

Output coding:
- Straight binary: 0 → 0 V, full‑scale code near Vref; best for unipolar inputs.
- Offset binary: Mid‑scale represents 0; convenient for AC‑coupled or pseudo‑bipolar ranges.
- Two’s complement: Negative and positive codes around mid‑scale; common for true bipolar inputs.
Framing and alignment: Choose MSB‑first or LSB‑first, left/right justification and any zero‑pad bits so words align with your processor.
Signalling levels: Single‑ended logic suits modest speeds; differential pairs improve noise immunity and timing margin at higher sample rates.
Handshakes: Watch DRDY/CLK edges and any BUSY/CS timing so acquisition, conversion and readout don’t overlap or corrupt codes.

Practical examples on microcontrollers (e.g., Arduino)

A quick, reliable way to see analogue‑to‑digital conversion in action is an Arduino reading a trimpot. Most Arduino‑class boards use a 10‑bit ADC, so analogRead() returns 0…1023. Convert the code to volts with your actual reference (Vref), which is commonly 5 V or 3.3 V depending on the board.

const float Vref = 5.0; // set to 3.3 on 3V3 boards or your chosen AREF
void setup(){ Serial.begin(115200); }
void loop(){
  int code = analogRead(A0);
  float volts = code * Vref / 1023.0;
  Serial.print("Code: "); Serial.print(code);
  Serial.print("  Volts: "); Serial.println(volts, 3);
  delay(10);
}

Set the reference: Use the cleanest Vref available; match the formula to 5 V or 3.3 V.
Filter and settle: Add a small RC low‑pass and keep source impedance modest so the sample‑and‑hold settles.
Average wisely: Sum 8–16 readings and divide to reduce noise; don’t violate Nyquist if the signal changes quickly.
Scale safely: Use a divider or buffer op‑amp to keep inputs within 0…Vref.

How to choose an ADC for your project

Start with the signal. Define its bandwidth (fmax), amplitude range, noise floor and whether it’s DC, AC‑coupled, single‑ended or differential. Then pick an architecture that matches: SAR for versatile, low‑latency data acquisition; delta‑sigma for high‑resolution, low‑bandwidth precision; pipeline or flash for very high speed. Finally, check the surrounding pieces—reference, analogue front‑end, filtering and the digital interface—because they often set the real‑world limit.

Bandwidth and rate: Choose fs ≥ 2 * fmax, add margin for filtering, and include an anti‑alias low‑pass.
Resolution and ENOB: Size n bits and required ENOB so LSB = Vref / 2^n resolves the smallest meaningful change.
Accuracy: Check INL/DNL, offset/gain error and drift against your tolerance.
Latency and jitter: Tight control loops prefer SAR; streaming/FFT work needs low clock jitter.
Inputs: Single‑ended vs differential, allowed common‑mode, input swing, required drive impedance.
Reference plan: Low‑noise, low‑drift Vref; ratiometric if the sensor allows.
Interface: SPI/I²C for MCU speeds; LVDS/CMOS parallel or JESD‑style links for high throughput.
Power and size: Confirm supply rails, consumption and thermal limits.
Integration: Consider on‑chip PGAs, multiplexers or digital filters to reduce BOM and complexity.

ADC choices for audio, video and film digitisation

The right analogue‑to‑digital conversion path for legacy media depends on signal bandwidth, required dynamic range and how much latency your workflow can tolerate. Getting this match right is what preserves quiet audio details, fine video texture and delicate film shadow tones without adding artefacts or losing sync.

Audio (tapes, vinyl, reels): Favour high‑resolution delta‑sigma ADCs for excellent linearity and low noise. Prioritise low THD, high SNR/ENOB and a clean clock to minimise jitter‑induced noise at higher audio frequencies. Latency from digital filtering is acceptable in transfers, so choose quality anti‑alias filters and a low‑noise Vref to maximise usable bits.
Analogue video capture (VHS, Betacam, etc.): Bandwidth spans into the MHz range, so prefer pipeline (or fast SAR) converters that offer high sample rates with fixed, predictable latency. Look for strong SFDR, low aperture jitter and good INL to control spurs and preserve fine luminance/chroma detail. Use a well‑designed low‑pass front‑end to contain aliasing.
Film scanning (8 mm, Super 8, 16 mm): Frame‑by‑frame workflows trade bandwidth for precision. Choose high‑linearity, high‑ENOB SAR or delta‑sigma devices; stable references and monotonic behaviour protect highlight/shadow gradients. Differential inputs and careful filtering help when reading very low‑level optical signals.

Best practices for reliable conversions

Reliable analogue‑to‑digital conversion is mostly won before the ADC pin: shape the spectrum, set levels, time the sample, and keep noise out. Treat the signal path, reference and clock as one system. Decide the bandwidth, obey Nyquist with margin, and design a front‑end that lets the sample‑and‑hold settle every time. Then lay out the board so digital edges can’t pollute sensitive nodes. These habits prevent most headaches in audio, video and measurement projects using analog to digital conversion.

Filter first: Add an anti‑alias low‑pass with cutoff comfortably below fs/2.
Stabilise Vref: Use a low‑noise, low‑drift reference; decouple right at REF pins; route quietly.
Tame jitter: Provide a low‑phase‑noise clock and clean supply; isolate it from fast digital lines.
Drive the S/H: Buffer and scale inputs; keep source impedance low; meet acquisition/settling time.
Use differential wisely: Prefer differential inputs for better CMRR; set the correct VCM.
Layout and grounds: Short, shielded input/reference runs; disciplined analogue/digital returns with a single‑point tie.
Protect the ADC: Series resistors and clamps/TVS to limit surges and out‑of‑range events.
Prove performance: Calibrate offset/gain; verify ENOB/SFDR with known tones; average/oversample judiciously without violating Nyquist.

Glossary of common ADC terms

Keep this quick reference handy when reading data sheets or tuning your analogue‑to‑digital conversion chain. These concise definitions match the concepts and specs covered above.

Sampling rate (fs): Conversions per second; fs = 1/T.
Nyquist frequency: fs/2; sample at least 2 × fmax.
Aliasing: Undersampling artefact; high frequencies fold low.
Resolution/bit depth (n): Distinct codes = 2^n.
LSB/step size: LSB = Vref / 2^n.
Quantisation noise: Rounding error from finite steps.
SNR: Signal‑to‑noise ratio, excluding distortion.
ENOB: Effective bits from SINAD; real usable resolution.
THD: Sum of harmonic distortion components.
SFDR: Gap between fundamental and largest spur.
INL/DNL: Transfer curve bowing/step uniformity; monotonicity.
Aperture jitter: Sampling‑instant uncertainty; worse at high f.
Latency/group delay: Time from sample to valid code.
Vref/full‑scale (FS): Reference sets FS and code size.
Sample‑and‑hold (S/H): Captures the instantaneous input for conversion.

Key takeaways

Successful analogue‑to‑digital conversion is about faithful capture, not just big numbers. Get the bandwidth and rate right, pick an architecture that fits, and keep the front‑end, reference and clock clean. Plan for latency and interface timing so the data you record is accurate, aligned and easy to use downstream.

Know the core: Sample → quantise → encode; obey fs >= 2 * fmax and anti‑alias.
Match resolution: Bits and Vref set LSB; size Vref to your signal swing.
Read the specs: SNR/ENOB for usable detail; INL/DNL and THD/SFDR for accuracy and purity.
Pick the right ADC: SAR for general purpose, delta‑sigma for precision, pipeline/flash for speed.
Mind timing: Acquisition, conversion latency and jitter can make or break performance.
Build a solid AFE: Buffer, filter, scale, protect; use a low‑noise, stable reference.
Stream cleanly: Choose output coding and bus framing that your host can capture reliably.

Digitising tapes, film or discs and want it done right? Our Canberra team handles end‑to‑end transfers in‑house. Start here: National Video Centre.